ferrite bead experiment

LT Spice thinks they can be. It lists a bunch of Wurth ferrite beads running from 16nH to 14uH (with a whole range of maximum currents and parallel and series resistances).

You can't rely on them being lossy enough to kill every resonance.

Yup. Try the inductor paralleled with a low-capacitance TVS, as well. The parallel RL plus shunt C is basically a one-pole network with low voltage drop, and that's a pity.

In our usual configuration the TVS charges up the reservoir cap without putting a lot of 1/2 LI**2 into the inductor to cause ringing, and then limits the ring amplitude if a load suddenly gets disconnected.

Antiparallel low-C TVSes across the inductor are good if you're running too close to somebody's abs max. You can use antiparallel rectifiers, but the have a lot of capacitance, which trashes the filter performance.

Very likely. Beads vary all over the place. Laird and Murata have decent info about saturation, which lots of other makers don't.

Cheers

Phil Hobbs

The bead application is mostly local noise filtering of the power to a pcb pour or a small subcircuit. That's usually downstream of a regulator so there are no big power transients to store a lot of energy.

A resistor and a cap are good if there's not a lot of current required. Like, say, 5 ohms and a 56 uF polymer cap. That won't resonate!

The data on Muratas site is good, and you can find PSpice models that are pretty good. If you do the test, my guess is that it will be very close to the simulation

Beads is often used to clamp 100MHz+ frequencies from transient edges from SMPS switching nodes, that otherwise will crop up on radiated emissions tests. I do not see why you would use a bead for reducing noise in a supply rail, the noise you see on a rail would be orders of magnitude below the lossy region of the bead, right?

Sounds like good fun. I goof around with filters from time to time. It's interesting to see what changing the core material does to the efficacy and the VNA I bought recently is the ideal tool for seeing what's going on in great detail. I recall - going back 50 years or so - there used to be an amazing choice of core materials; astonishing range of differing grades of powdered iron, ferrite, 'feroxcube' (whatever that was) and various other exotic materials. I've still got a big pile of 'em somewhere I can't find. But I'm guessing that amazing range is a thing of the past, given the frequencies designers are messing with nowadays. Would that be a correct assumption?

I have been doing some consulting, so now I can afford new equipment. Next week I am buying the Siglent SVA1015X, with the EMC pack and a LISN, so I can do conducted emission measurements in my own lab

Next is the Bode 100, but that only goes to 50MHz, but it's a really nice tool with great SW package. Very high price of 5000 USD compared to whats actually inside the box, in this case on is paying for the SW

Sure, I do the same thing, though generally with a ceramic in order to save space. But diodes vs. resistors for de-Qing are very often a win.

Cheers

Phil Hobbs

There is a wide range of materials. In the specific case of, say, some various 1206 ferrite beads that are specified for 75 ohms at 100 MHz, there is radical variation in low frequency inductance and saturation behavior, which is not often well documented.

The idea of using beads instead of inductors to kill the Q of a power supply lowpass filter sounds good, but may not work. I guess it usually doesn't.

DC saturation complicates that too. May as well just try it.

One issue we have is ripple and spikes from switching supplies getting into analog stuff and fast jitter-sensitive things. The switchers might run at a MHz or so, but the spikes go way up.

This is usually deep inside the circuits, so it's not a radiated EMI concern. We could have a dozen supply rails to filter on a board.

Beads are specified by their Z@100 MHz, so they're almost always inductive below there.

Cheers

Phil Hobbs

So they are just cheap inductors at the frequencies of concern. Maybe they need a resistor in parallel.

We'll experiment.

They are cheap, non-wound inductors so the parallel capacitance is low. Resistors in series - or parallel - can sometimes be useful.

You could try thinking first. LTSpice does let you simulate beads/chips - each one of the devices they let you pick has it's own Spice model with parameter values.

It's quicker than soldering stuff together, and the test gear is much cheaper.

Yes. The Murata Simsurfing tool shows a lot of detail:

Not a bad idea and soemtimes quicker than theoretical methods. Let's just confirm that it works in practice.

In Europe, Philips / Mullard made a vast range of ferrite materials back in the 60's / 70's, which were very well documented in terms of application. Still have some of the data sheets and handbooks somewhere and we used them extensively for small converters and filter applications. Even RF power amplifiers by stacking pair sets between pcb end plates...

Chris

If you happen to have a link for muRata saturation performance, would you mind posting it? I've looked before, but did not find it. Wurth does publish some info on impedance versus current.

I have a rule of "no more than 10% rated current if you want anything close to low frequency inductive performance seen in the plots." There is a remaining question of impedance (really, they are resistive) at higher current and higher frequencies. Is the resistive performance at higher frequencies degraded in the same fashion due to DC current, or is it less so?

It isn't easy to measure. I've done some implicit de-embedding for FB bias for MMIC amps. I take the simulated-modeled frequency response for 0 A through the FB in an MMIC amp app, and then go measure the frequency response of an actual amp with DC bias using that bead. Then I degrade the FB model in the simulator till it is a decent match to the measured response. It isn't very accurate, and is very case specific, but it is better than nothing.

Put two beads in series for DC and in parallel for AC. Measure the impedance of the junction vs current. Multiply by 2.

Am 08.05.21 um 00:01 schrieb John Larkin:

The VNA usually has a built-in bias tee for each port that is already calibrated in. So just insert the bias current into the backside BNC of your VNA and that's it.

Gerhard

I don't have a VNA; we work all time domain here. Lots of nonlinear stuff.

I'd measure inductance by connecting a 50 ohm signal generator and finding the 3db frequency, maybe prowling around for resonances and such.

What's the tee like on a typical VNA? The dual-inductor trick avoids caring about the tee. If the inductor measurement needs 15 amps, no problem.

It would be a pity to end up measuring the saturation performance of the bias tee in the VNA. John's method is more elegant, provided you have low-enough-impedance decoupling on the end of the bead where you connect the power supply. At least you can measure that fairly well to confirm the assumption.